The Behavior Score
Behavior score / noun / — a measure of whether an account's usage looks like a legitimate customer journey. A behavior score comes from behavioral scoring of ordinary product activity (navigation shape, event timing, session structure, and repetition) compared against how genuine users actually behave.
Dregs computes a Behavior score from 0 to 100 for every user, continuously, as one of its four identity scores, using the event stream you already send. It can catch abusers the other dimensions can't: even real humans, with plausible identities, on clean devices, might be doing something wrong.
The Behavioral Analysis Behind the Behavior Score
Real users wander, pause, backtrack, and explore. Abusers execute. The Behavior score quantifies the difference:
How Dregs Computes the Behavior Score
A pipeline of analyzers runs over each identity's event history, each producing observations with a value from 0.0 (suspicious) to 1.0 (legitimate), a confidence, and an explanation. The Behavior score is the weighted aggregate of the Behavior-category observations, scaled to 0–100 and recomputed as new events arrive, so it tracks what the account is doing now, not what it did at signup.
{
"category": "BEHAVIOR",
"id": "behavior.event-timing",
"label": "Event Timing",
"value": 0.15,
"confidence": 0.8,
"explanation": "Event cadence is machine-regular across sessions",
"metadata": { ... }
}Behavioral Scoring Catches What Signup Checks Can't
Every other fraud signal is strongest at signup and decays from there. Behavioral scoring is the opposite: it starts thin and compounds, because behavior is the one thing an abuser has to keep doing. A trial cycler can fabricate a new identity in thirty seconds, but they can't stop navigating like someone who's done this five times. A scraper can rotate devices, but it can't stop moving like a scraper.
That's why the Behavior score pairs so well with the signup-time dimensions, and why it's central to the free trial abuse, credential stuffing, and bot detection use cases. For the discipline behind the score (the signals, the baseline, the comparison to UBA and UEBA tooling) see behavioral analytics.
Not the Credit-Bureau Behavior Score
If you arrived here from the lending world: consumer credit has its own "behavioral score," computed from repayment history to predict default risk. This page is about the other one: behavioral scoring of how users act inside web applications, for fraud detection and abuse prevention. Same phrase, different universe.
The Other Three Scores
Behavior answers "does the usage make sense?" The other three dimensions profile who's doing the using.
Humanity
Is a real human controlling the browser, or automation?
Authenticity
Does their claimed identity look real, or fabricated?
Uniqueness
Is this the user's only account, or one of many?
All four come from the same identity scoring pipeline, on every plan, from the events you're already sending.
Frequently Asked Questions
Q: What is a behavior score?
A: A behavior score is a continuous measure of whether an account's activity looks like a legitimate user's. In Dregs it's a 0–100 value computed from how the account actually uses your product (navigation shape, event timing, session structure, and repetition) compared against the patterns of real customers. A score near 100 moves like a genuine user; a score near 0 moves like a script, a scraper, or someone executing a routine they've run many times before.
Q: Is this the same as the credit-bureau "behavior score"?
A: No. In consumer lending, a behavioral score is a credit-risk number computed from repayment and account-management history. Dregs's Behavior score has nothing to do with credit: it's behavioral scoring for fraud and abuse detection in web applications, measuring whether an account uses your product the way a real customer does. Same phrase, entirely different industry.
Q: What's the difference between the Behavior score and behavioral analytics?
A: Behavioral analytics is the discipline of analyzing activity patterns over time. The Behavior score is the product of applying it: one continuous 0–100 number per identity that you can sort, badge, alert, and automate on. Dregs's behavioral analytics page covers the discipline and the signals in depth; this page covers the score those signals roll up into.
Q: What behavioral scoring signals work without labeled training data?
A: All of Dregs's out-of-the-box signals: navigation efficiency (first-time users wander; abusers and scripts take the shortest path, every time), event timing (human pauses versus machine cadence or scheduled-looking intervals), session structure (depth, length, time-of-day spread), and repetition across sessions and across linked accounts. Nothing needs labeling: the signals come from the events you already send.
Q: How is the Behavior score different from the Humanity score?
A: Humanity asks whether a human is present at all, based on automation signatures, device rendering, and inhuman timing. Behavior asks whether the usage makes sense as a customer journey, which catches humans behaving abusively: the freeloader who navigates your trial with surgical precision because it's their fifth account, the credential tester working through a list, the human-driven scraper. Low Humanity means "probably a bot"; low Behavior with high Humanity means "a real person, but up to something."
Q: How quickly does behavioral scoring become useful?
A: Within-session signals (timing, navigation shape, repetition) are useful from the first day. Signals that compare an account against your real-customer baseline sharpen as Dregs accumulates data over the following weeks. Confidence is reported with every observation, so early scores are honest about how much evidence sits behind them.
Behavioral scoring from the events you already send.
Drop the Dregs tracking script into your application and every identity gets a continuous Behavior score (plus Humanity, Authenticity, and Uniqueness) with no labeling and no training period.
Schedule a Demo